CosmoFit.EBOSSELGFullShapeLikelihood

class CosmoFit.EBOSSELGFullShapeLikelihood(cosmology, version='dr16')

Bases: TabulatedBAOLikelihood

eBOSS DR16 emission-line galaxies, full shape: the joint likelihood of (D_M/r_d, D_H/r_d, f sigma_8) at z_eff = 0.845, on a 100x100x100 grid.

The same galaxies as EBOSSELGLikelihood, analysed differently. That one compresses the clustering to a single isotropic BAO scale; this one keeps the anisotropy and the growth rate as well, so it constrains the amplitude of structure directly rather than only the geometry.

Being three-dimensional is the point. f sigma_8 is degenerate with the Alcock-Paczynski distortion, and the grid holds that degeneracy exactly, which two separate error bars could not. It is also why the prediction here uses fsigma8(z) unrescaled: the geometry the measurement was made against is a coordinate of the grid, not a fiducial to correct back to, so applying the AP correction that FSigma8Likelihood needs would count it twice.

Warning

Mutually exclusive with "eboss_elg" – the same galaxies, twice. Overlaps "desi", whose ELG sample succeeds this one. It also overlaps "fsigma8", whose compilation includes eBOSS growth-rate measurements.

Parameters:

version (str)

__init__(cosmology, version='dr16')

Parameters:

version (str)

Methods

__init__(cosmology[, version])
chi2()	-2 log L, so this composes with the Gaussian likelihoods it is summed with.
log_likelihood()	Log-likelihood at the current cosmology.
log_likelihood_at(values)	Interpolate the released surface at given distance ratios, with no cosmology involved.
model()	Predicted distance ratios, one per tabulated observable.
predictions()	Alias for model().
residuals()	Prediction minus the grid's most probable point.
summary()	Return a summary of the likelihood evaluation.

Attributes

FAMILY	Loader family, set by the subclass.
LABEL	Display name, set by the subclass.
n_data	Number of data points.
name_and_size	Name together with the number of data points.